Sunil Venkataram
Back to garden
•
🌰 seedling
Distribution as the Remaining Moat - Why SaaS Incumbents Aren't Dead
type/notesaasaimarket-structure

Distribution as the Remaining Moat — Why SaaS Incumbents Aren’t Dead

The argument

Balaji Srinivasan in a 2026 interview, pushing back on the common SaaS-apocalypse framing:

Cloning all of Facebook’s code and spinning up a competing domain does not get you users or ad rates. AI accelerates both incumbents and disruptors equally, so the relative threat is lower than headlines suggest.

The claim has three parts:

1. Distribution is a separate asset from code. Every successful SaaS company has two things: a codebase and a user base. The codebase can now be cloned by AI at trivial cost — paste the public functionality into Claude or GPT and get most of it back in a few hours. The user base cannot. It took years to build, it requires trust and habit, and it sits in the heads of customers who have integrated the product into their daily workflow.

2. AI is a symmetric productivity multiplier. Native AI challengers use AI to build fast. Incumbents also use AI to build fast. If both sides get the same 10x speedup, the relative gap between them doesn’t change — it’s still the same user-base-to-challenger-base ratio that existed before AI. The incumbent’s position is preserved in percentage terms, even if both sides accelerate in absolute terms.

3. Distribution is the hardest thing to replicate. Network effects, switching costs, data lock-in, brand trust, mental habit, API integrations, training investments, procurement relationships — all of these are distribution assets that a new entrant has to rebuild from zero. AI doesn’t help with any of them directly. A startup that can now produce a Figma-quality tool in a weekend still has to acquire Figma-quality users, which is the part that takes years.

What this disagrees with

This argument is in direct tension with several claims held elsewhere in this knowledge graph. The tension is honest and worth naming explicitly.

System of Action - Evolution Beyond System of Record argues that native AI challengers flip the control relationship with incumbents by capturing Hero users and forcing integrate-and-surround. That atomic predicts incumbent control-point software loses the strategic high ground. Balaji’s argument pushes against this: the Tidemark frame is overstated because it doesn’t weight distribution heavily enough as a persistent moat.

Grow 10 or Earn 40 - Two Paths for Mature Software Companies argues that the middle of mature software disappears — either accelerate growth or rebuild for fortress margins. Balaji would agree that bad incumbents get compressed (he names NetSuite) but disagree that distribution-strong incumbents face the same pressure.

Run the Business vs Do the Work - The AI-Era Vertical SaaS Shift argues that historical vertical SaaS (run the business) gives way to AI-native vertical SaaS (do the work). Balaji would say this is true for some verticals but not universal — distribution-strong incumbents can ship their own “do the work” products fast enough, and the population of users they already have is a structural advantage.

The reconciliation

Both positions are serious. The reconciliation depends on splitting SaaS incumbents into two categories:

Category A — Healthy incumbents with real distribution.

  • Still shipping improvements that users value
  • User base is growing or stable, not milking
  • Engineering culture can absorb AI tools and ship fast
  • Brand trust and mental habit are intact
  • Distribution is a genuine moat

For this category, Balaji’s argument holds. The distribution moat is real, AI accelerates both sides equally, and the incumbent can absorb the transition without losing the position. Examples: Figma, Linear, Notion, Vercel, Shopify in its main verticals.

Category B — Complacent incumbents milking installed bases.

  • Product has been stagnant for years
  • Engineering velocity is slow
  • User base is retained by switching cost, not by value delivered
  • Culture resists PLG, resists shipping, resists change
  • Distribution is a lagging indicator, not a forward asset

For this category, the Tidemark/David George/Luminai arguments hold. Distribution that exists because customers have no alternative is not really distribution — it’s a rental agreement that expires when a good alternative shows up. Balaji explicitly names NetSuite as an example.

The question for any specific SaaS company is: which category? Investors and operators should run the diagnostic on any holding or pitch:

  • Is the product still improving at a pace customers notice?
  • Is the user base growing on merit or retained on inertia?
  • Can the engineering org ship an AI-native feature in weeks, not quarters?
  • Would customers stay if a native AI alternative were obviously better?
  • Is distribution a forward asset (compounding) or a lagging asset (decaying)?

Category A companies have a durable moat and Balaji is right about them. Category B companies are on the conveyor belt and the convergence thesis is right about them. The universal form of either claim is wrong — reality bifurcates along these diagnostic lines.

Why the bifurcation matters

Tonight’s knowledge graph accumulated a five-source convergence thesis (Karpathy, David George, Kesava/Luminai, Tidemark, Tanay Jaipuria) around “value migrates to the outcome/action layer, not the record layer.” That thesis is still correct as a description of what is happening to Category B incumbents. It is overstated as a universal prediction. Balaji is the honest 6th source that complicates the picture — he agrees with the others on some verticals and pushes back on others.

The practical implication: don’t short well-distributed SaaS indiscriminately. Short the Category B bucket. Hold or even long the Category A bucket if valuations are reasonable. The binary framing (“SaaS is dead” or “SaaS is fine”) loses money. The bifurcated framing sorts the winners from the losers.

What would make Balaji wrong

Three ways the distribution moat could fail even for Category A incumbents:

1. AI becomes so good that user habits reset. If AI agents become the primary consumers of SaaS products (see Claws - Persistent Looping Agents as App Replacement), the user-level habits and relationships that constitute distribution may become irrelevant. An agent doesn’t have brand loyalty. An agent selects products by API quality and cost.

2. AI tooling lowers the cost of rebuilding distribution itself. If AI-assisted GTM motions can compress the time-to-build-distribution from years to months, the historical distribution moat shrinks as an asset. This is partially happening already with PLG at scale, but not yet with enterprise distribution.

3. Regulatory or interop forces break the data lock-in. Open banking, open EHR standards, and similar interop requirements can turn accumulated data from a moat into a liability. Category A incumbents whose moats depend heavily on data lock-in are vulnerable if this happens.

Balaji’s framing is the default, not the guarantee. Watch these three forces for signals that the distribution moat is eroding. So far, only the first is showing meaningful early signs.

Backlinks (4)
Connected Notes